On Fri, Jun 06, 2025 at 10:21:16 +0100, Daniel P. Berrangé wrote:
> On Fri, Jun 06, 2025 at 11:05:23AM +0200, Peter Krempa wrote:
> > On Fri, Jun 06, 2025 at 09:52:49 +0100, Daniel P. Berrangé via Devel wrote:
> > > From: Daniel P. Berrangé <berra...@redhat.com>
> > >
> > > Bug reports from automated tools and AI agents are time consuming to
> > > triage and have poor signal/noise ratio. Set strong expectations for
> > > any reporters using such tools, in a (likely doomed) attempt to stem
^^^^ [1]
> > > the flow of poor quality reports.
[...]
> > > +Use of automated tools / AI agents
> > > +----------------------------------
> > > +
> > > +If any automated tool / AI agent is used to identify a bug / security
> > > +flaw, the following additional expectations apply when filing a report:
> > > +
> > > +- The tool / agent used **MUST** be clearly declared in the description
> > > +- All stated facts **MUST** be validated as correct and free from AI
> > > + hallucinations prior to filing
> > > +- The problem **MUST** be described against an upstream release that is
> > > + no more than 3 months old.
> > > +- The problem **SHOULD** be analysed and accompanied with a proposed
> > > + patch that can be directly applied to current git
> >
> > I'd also like to prohibit/avoid vague and too general statements.
> > In the few last reports that were low quality that I've seen, the
> > problem statement and reproducer were true because they were too vague.
> >
> > E.g. saying that "if you call this function with NULL argument it will
> > crash" can be true, but if we're making sure that it can't happen
> > elsewhere it's quite useless.
> >
> > I'm not sure though how to formulate that.
>
> I figure that kind of vague / have-wavy nonsense is often a characteristic
> of AI output. By requiring use of AI to be declared upfront, when we see
> such vague statements, we can just dismiss the bug or require the reporter
> to explain properly.
Good point. Also as you point out [1] it's likely that slop submitters
won't conform to this either; mostly because they'd have to put effort
in reading this which goes against the use of slop generators.
